I really enjoyed this! I would love hearing more stories about deploying Go apps to production. I have really enjoyed using Go so far, but haven’t deployed anything that needs to handle large amounts of traffic.
I really enjoyed this! I would love hearing more stories about deploying Go apps to production. I have really enjoyed using Go so far, but haven’t deployed anything that needs to handle large amounts of traffic.
FWIW, my team recently found a hotspot in our Python code – a function that was called once per raw record in a multi-terabyte data set we had to reprocess frequently. We decided to bake off some approaches to optimizing the hot spot. (The slow function was urllib.urlparse.)
The three approaches chosen:
Our results were that the Go program was about 5X faster than the CPython version. But the PyPy version was 3X faster and the Cython version with an optimized rewrite of that function was actually 10X faster.
YMMV, but to us, if PyPy is within striking distance to Go and Cython is available as an easy optimization option, then Go isn’t worth it – that is, if your only criterion is raw CPU performance.
There is some additional information from the talk author sprinkled around in the discussion on this over at HN: https://news.ycombinator.com/item?id=10402307